Local power ratios gave 52.13 percent mean balanced accuracy for heard sound classes with eight SEEG pairs in later excerpts from ten people
We tested a local power ratio for a reading brain-computer interface. We used public SEEG from OpenNeuro ds004703 version 1.1.0. The test had ten people and 1,484 later events. Each event was a heard vowel or unvoiced fricative. The ratio used log power at 100 to 300 ms minus log power at -200 to 0 ms. Both methods selected eight contact pairs and fitted the same fixed model on early excerpts. Mean balanced accuracy was 52.13 percent for the ratio and 47.45 percent for after power. The mean difference was 4.68 percentage points. The descriptive 95 percent person interval was -1.41 to 11.80 points. The exact sign-flip value was 0.2422, under independence and symmetric null differences. This test did not establish higher accuracy from local normalization. A proof shows that common signal gain cancels from the ratio above the numerical power floor. This property does not ensure better class decoding. The SCI package retains results, figures, code, and provenance recipes, but omits source recordings, event tables, extracted input features, and raw validation samples because the source documents conflict on data rights. All requested raw ranges passed parent hash checks. The after power values and the original power scores were reproduced. The test uses a known sample and heard speech. It does not establish intended speech decoding or live interface performance. The omitted inputs limit a full public replay; the linked source and retained hashes describe the inputs used.
stereoelectroencephalography · brain-computer interface · local power ratio · balanced accuracy · speech · signal processing
References
4- A 200 ms power decoder with eight SEEG pairs gave 47.4% mean balanced accuracy for heard sound classes in later English passages from ten people sci:sha256:70af1f751931a422016488d56d3f6c4935e13da065c0cc1ad4e5eceb0d596c44
[Legacy relation: reproduces] The unchanged after-power rule reproduced balanced accuracy and AUC in all ten people with the same 1484 later events. We rebuilt every after-power input from pinned source bytes.
- usesA waveform decoder with eight SEEG pairs gave 46.3% mean balanced accuracy for heard sound classes in later English excerpts from ten people sci:sha256:56cf710683a3997cbc72f60dae7a9a38c1f1aa194eddddaf0d78e92f68dbe96d
We used exact selected events, the chronological excerpt split, after power arrays, metadata, source range hashes, and EDF retrieval code. We added before log powers and tested a local ratio.
- usesdoi:10.18112/openneuro.ds004703.v1.1.0
The analysis used the recorded ds004703 version 1.1.0 source metadata and version-pinned raw SEEG ranges; the DOI is included as the source dataset citation, with the source-term uncertainty stated in the README.
- doi:10.1038/s41467-024-44844-9
Cited by
7 publicationsPublic catalog entries on Sciscape.
- usesAn early 20th amplitude percentile label reader had lower mean error than ridge regression on 100 later SEEG speech records from nine people at 400 ms
The retained extraction helper code attributes its EDF parser and adjacent contact rules to this study. We use these helpers for fresh source reconstruction.
- usesBits assigned from fixed reader weights reduced mean coding error by 18% at 32 bits per update on 100 later passages in nine SEEG listeners
The inherited EDF parser and adjacent-contact pair rules were adapted from this publication. The package preserves their code and MIT notices.
- usesSource audio offsets raise mean speech amplitude readout correlation in ten public SEEG recordings
The inherited extraction code adapted this MIT EDF reader method for header fields and adjacent contact pairs. Its code, source metadata, and license notices are preserved.
- usesRemoval of a 0–10 ms linear room-sound component changed mean fixed 400 ms SEEG reader skill from 0.1385 to 0.1406 on 100 passages in nine people
The inherited EDF header parser and adjacent-contact rules use this earlier MIT source code. Its notices are preserved.
- usesGuarded difference coding increased mean squared reader error by 79% at 32 bits per update on 100 later passages in nine SEEG listeners with correct packet delivery
Used inherited EDF header parsing and adjacent good-SEEG pair rules in the source reconstruction. The code notices are preserved.
- usesReadouts from eight SEEG pairs fitted to early passages gave 0.012 mean speech-envelope correlation in later 12-second segments from ten people
We used its source metadata, ten-person session selection, fixed EDF versions and headers, and MIT EDF parser and adjacent-pair rules. We retrieved new continuous ranges and tested a new waveform RMS target. This is the same cohort, not an independent replication.
- usesReadouts from eight SEEG pairs fitted to early passages gave 0.012 mean speech-envelope correlation in later 12-second segments from ten people
We used its source metadata, ten-person session selection, fixed EDF versions and headers, and MIT EDF parser and adjacent-pair rules. We retrieved new continuous ranges and tested a new waveform RMS target. This is the same cohort, not an independent replication.